The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
206
does it require the orthoimagery at high resolution. Even
parameters such as mosaic seamlines are not dependant on the
requirement for orthoimage with final accuracy, Most of the
dataflow and production bottlenecks can be alleviated by using
geodatabase and server centric workflows.
2. INTRODUCTION TO ARCGIS IMAGE SERVER
ArcGIS Image Server is a product from ESRI that provides
management, processing, and dissemination of very large
volumes of imagery with on-the-fly server based processing.
This product can serve traditional pre-processed image tiles,
and serve imagery that has not be pre-processed. The required
processing is applied in real time as the imagery is being served.
The processing to be applied is defined in a database. This
enables the same imagery to be served in different forms, as
required by different users, without additional data storage.
Also, the imagery can remain in its native or original form
removing the requirement to convert the data. Additionally, the
image products can be served as virtual mosaics. When an
image is requested, the server not only extracts the required
pixels, but applies the required processes, returning the pixels as
well as the metadata to the user’s application.
As with the other ArcGIS Server family of products, the
workflow is defined in three stages: Author, Serve, and Use.
2.1 Author
In authoring an image service definition database is created,
which defines a catalog of all the imagery. This image service
definition contains information on each of the individual images
including all properties, metadata, footprint, and location. The
large volume of pixel data are not loaded into the database, but
remain in their original form and are referenced. Also,
associated with each image are the properties of how the images
should be processed. These include the processes to be applied
and the parameters of the processes. Such processing can
include how to radiometrically enhance the images as well as
how to process the image geometrically. A geometric process
for georeferencing can be a simple affine transform or more
complex transforms such as orthorectification that is dependant
on camera calibration information, orientation data, and a
reference to an elevation model. Associated with each image is
a footprint shape that defines the extent of the pixels to be used,
enabling the exclusion of pixels that represent the data frame of
a scanned image or exclusion of areas such as clouds. The same
set of images can be included in different image sendee
definitions enabling different imagery products to be defined
with no additional imagery being stored. Such products could
include, different band combinations or enhancements that
utilize the multispectral capabilities of many digital sensors.
Central to ArcGIS Image Server is the image service definition
that defines within a database structure the processing to be
performed on the imagery as well as all the parameters that
effect the processing. As processes and parameters become
available in the production flow they can be added or modified
within the image service. The service definition consists of both
processes defined using XML as well as the parameters of these
processes defined either within the XML or as attributes of the
service tables. This provides great flexibility and enables the
database table to define only those parameters that may change.
For example, parameters such as interior orientation that do not
change can be defined as constants in the XML while
parameters such as the six external orientation parameters can
be defined in the tables. These orientation parameters may be
defined initially from the GPS and flight navigation data and be
updated when the GPS/IMU parameters are available, then
updated after an aerial triangulation is performed.
The open and documented data structures used enable simple
integration of these parameters from a wide range of existing
applications that perform such parameter determination. This
enables ArcGIS Image Server to be quickly integrated into
existing workflows.
2.2 Serve
The image service definition can be published on the server and
then is simultaneously accessible to many different client
applications. These applications can either connect directly to
ArcGIS Image Server using RPC (Remote Procedure Calls) or
through standards such as OGC WMS, WCS as well as SOAP
and REST by serving through ArcGIS Server. These enable a
large range of client applications to access the imagery as
services, either over local area networks or over the web.
2.3 Use
Clients connecting to image services can view the imagery as a
large, single image; although they often consist of thousands of
individual images. Importantly, metadata about the image
service as well as the individual images being viewed can be
accessed, providing the client with information about the
imagery that is critical to many decision making tasks. The
client applications can also interact with the image services to
dynamically refine the processing. For example, a client may
change the order of overlapping imagery to obtain imagery that
more closely matches their requirement, such a being closer to a
specified date or having a required sun elevation. When
working over low bandwidth connections, a client can change
the compression used for the data transmission. The
compression used for the data transmission is therefore
independent of the compression used to store the data, enabling
an analysis, for example, to use highly compressed imagery for
faster navigation, while still being able to reduce or remove the
compression for detailed analysis. Therefore, the required
image product can be created on demand, based on parameters
defined in the database or specified by the client application.
3. PROCESSES
For all requests, the multiple geometric processes such as
orthorectification, reprojection, pan-sharpening, and mosaicking
are concatenated internally by ArcGIS Image Server to a single
sampling of the original imagery. As a result, the quality of the
resulting pixel data is improved. Prior to sampling the imagery
the geometry is analysed such the only the pixels required for
processing are read, therefore, the volume of data to be read
and processed is substantially reduced. Typically, with aerial
imagery being flown with high overlaps only a small
percentage of the original pixels may be needed in the final
product and require processing. Performing pan-sharpening on-
the-fly without the need to store the large pan-sharpened file
also substantially reduces data volumes.